Install namd

• Donwload source code with the link.

• Unpack namd package: tar xzvf NAMD_x.xx_Source

• Unpack charm package:

  cd NAMD_x.xx_Source
  tar xvf char-x.xx.x
  cd char-x.xx.x

• Set the path for CUDA Toolkit with the exact path (mine is /usr/local/cuda-10.0): export CUDATOOLKIT_HOME=/usr/local/cuda-10.0

• Build charm: ./build charm++ multicore-linux-x86_64 cuda -j16 --with-production

• Check out the notes.txt (Examples givn in the next few bullet points)

  • Search Compiling NAMD

  • Check the extact name of multicore-linux-x86_64* directory. (Mine is multicore-linux-x86_64-cuda after building it with CUDA support. )

  • Test pgm (Mine has the -cuda postfix, which is not mentioned in the notes.txt file.)

    cd multicore-linux-x86_64-cuda/tests/charm++/megatest
    make pgm
    ./pgm +p4   (multicore does not support multiple nodes)
    cd ../../../../..

• Return NAMD_x.xx_Source directory

• Set CHARMBASE = .rootdir/charm-6.10.2 in Make.charm

• ./config Linux-x86_64-g++ --charm-arch multicore-linux-x86_64-cuda

• cd Linux-x86_64-g++

• make -j16

namd binary would be built at the end of the compilation.

• Copy binaries to your local bin directory (mine is $HOME/.local/bin)

  cp namd2        $HOME/.local/bin/
  cp charmrun     $HOME/.local/bin/
  cp flipbinpdb   $HOME/.local/bin/
  cp flipdcd      $HOME/.local/bin/
  cp psfgen       $HOME/.local/bin/
  cp sortreplicas $HOME/.local/bin/